[PATCH bpf] xsk: fix number of pinned pages/umem size discrepancy

From: Björn Töpel <redacted>

For AF_XDP sockets, there was a discrepancy between the number of of
pinned pages and the size of the umem region.

The size of the umem region is used to validate the AF_XDP descriptor
addresses. The logic that pinned the pages covered by the region only
took whole pages into consideration, creating a mismatch between the
size and pinned pages. A user could then pass AF_XDP addresses outside
the range of pinned pages, but still within the size of the region,
crashing the kernel.

This change correctly calculates the number of pages to be
pinned. Further, the size check for the aligned mode is
simplified. Now the code simply checks if the size is divisible by the
chunk size.

Fixes: bbff2f321a86 ("xsk: new descriptor addressing scheme")
Reported-by: Ciara Loftus <redacted>
Signed-off-by: Björn Töpel <redacted>
---
 net/xdp/xdp_umem.c | 17 ++++++++---------
 1 file changed, 8 insertions(+), 9 deletions(-)

diff --git a/net/xdp/xdp_umem.c b/net/xdp/xdp_umem.c
index e97db37354e4..b010bfde0149 100644
--- a/net/xdp/xdp_umem.c
+++ b/net/xdp/xdp_umem.c

@@ -303,10 +303,10 @@ static int xdp_umem_account_pages(struct xdp_umem *umem)
 
 static int xdp_umem_reg(struct xdp_umem *umem, struct xdp_umem_reg *mr)
 {
+	u32 npgs_rem, chunk_size = mr->chunk_size, headroom = mr->headroom;
 	bool unaligned_chunks = mr->flags & XDP_UMEM_UNALIGNED_CHUNK_FLAG;
-	u32 chunk_size = mr->chunk_size, headroom = mr->headroom;
 	u64 npgs, addr = mr->addr, size = mr->len;
-	unsigned int chunks, chunks_per_page;
+	unsigned int chunks, chunks_rem;
 	int err;
 
 	if (chunk_size < XDP_UMEM_MIN_CHUNK_SIZE || chunk_size > PAGE_SIZE) {

@@ -336,19 +336,18 @@ static int xdp_umem_reg(struct xdp_umem *umem, struct xdp_umem_reg *mr)
 	if ((addr + size) < addr)
 		return -EINVAL;
 
-	npgs = size >> PAGE_SHIFT;
+	npgs = div_u64_rem(size, PAGE_SIZE, &npgs_rem);
+	if (npgs_rem)
+		npgs++;
 	if (npgs > U32_MAX)
 		return -EINVAL;
 
-	chunks = (unsigned int)div_u64(size, chunk_size);
+	chunks = (unsigned int)div_u64_rem(size, chunk_size, &chunks_rem);
 	if (chunks == 0)
 		return -EINVAL;
 
-	if (!unaligned_chunks) {
-		chunks_per_page = PAGE_SIZE / chunk_size;
-		if (chunks < chunks_per_page || chunks % chunks_per_page)
-			return -EINVAL;
-	}
+	if (!unaligned_chunks && chunks_rem)
+		return -EINVAL;
 
 	if (headroom >= chunk_size - XDP_PACKET_HEADROOM)
 		return -EINVAL;

base-commit: 746f534a4809e07f427f7d13d10f3a6a9641e5c3
-- 
2.25.1

From: Björn Töpel <redacted>

For AF_XDP sockets, there was a discrepancy between the number of of
pinned pages and the size of the umem region.

The size of the umem region is used to validate the AF_XDP descriptor
addresses. The logic that pinned the pages covered by the region only
took whole pages into consideration, creating a mismatch between the
size and pinned pages. A user could then pass AF_XDP addresses outside
the range of pinned pages, but still within the size of the region,
crashing the kernel.

This change correctly calculates the number of pages to be
pinned. Further, the size check for the aligned mode is
simplified. Now the code simply checks if the size is divisible by the
chunk size.

Fixes: bbff2f321a86 ("xsk: new descriptor addressing scheme")
Reported-by: Ciara Loftus <redacted>
Signed-off-by: Björn Töpel <redacted>

Thanks for the patch Björn.

Tested-by: Ciara Loftus <redacted>

On Thu, Sep 10, 2020 at 2:29 AM Loftus, Ciara [off-list ref] wrote:

quoted

From: Björn Töpel <redacted>

For AF_XDP sockets, there was a discrepancy between the number of of
pinned pages and the size of the umem region.

The size of the umem region is used to validate the AF_XDP descriptor
addresses. The logic that pinned the pages covered by the region only
took whole pages into consideration, creating a mismatch between the
size and pinned pages. A user could then pass AF_XDP addresses outside
the range of pinned pages, but still within the size of the region,
crashing the kernel.

This change correctly calculates the number of pages to be
pinned. Further, the size check for the aligned mode is
simplified. Now the code simply checks if the size is divisible by the
chunk size.

Fixes: bbff2f321a86 ("xsk: new descriptor addressing scheme")
Reported-by: Ciara Loftus <redacted>
Signed-off-by: Björn Töpel <redacted>

Thanks for the patch Björn.

Tested-by: Ciara Loftus <redacted>

Acked-by: Song Liu <redacted>

On Thu, Sep 10, 2020 at 12:56 AM Björn Töpel [off-list ref] wrote:

From: Björn Töpel <redacted>

For AF_XDP sockets, there was a discrepancy between the number of of
pinned pages and the size of the umem region.

The size of the umem region is used to validate the AF_XDP descriptor
addresses. The logic that pinned the pages covered by the region only
took whole pages into consideration, creating a mismatch between the
size and pinned pages. A user could then pass AF_XDP addresses outside
the range of pinned pages, but still within the size of the region,
crashing the kernel.

This change correctly calculates the number of pages to be
pinned. Further, the size check for the aligned mode is
simplified. Now the code simply checks if the size is divisible by the
chunk size.

Fixes: bbff2f321a86 ("xsk: new descriptor addressing scheme")
Reported-by: Ciara Loftus <redacted>
Signed-off-by: Björn Töpel <redacted>

Applied. Thanks